Article URL: https://www.constraintlab.com/blog/ai-is-a-bubble-just-like-dot-com.html Comments URL: https://news.ycombinator.com/item?id=49185880 Points: 6 # Comments: 0

Andrej Karpathy has written large language models the way most of us have written CRUD apps. He was a founding member of OpenAI, ran AI at Tesla, and joined Anthropic this spring to train Claude. He wrote nanoGPT and llm.c, the small readable codebases people use to learn how a language model works end to end. When Anthropic put Claude inside Slack, he called it “a new paradigm” — the third major redesign of how humans use these systems. Not a website you visit anymore, not an app you install, but “a self-contained, persistent, asynchronous entity with org-wide tools and context, working alongside teams of humans.” Hold that exchange still for a second. Nobody in it is stupid. That’s what makes it worth understanding. The same split runs through everything right now, and nothing around it holds still. Leadership says “use more AI,” and never defines responsible use. Engineers hear it through an older instinct: don’t ship what you don’t understand. Nobody explains how both survive. Layoff announcements credit AI in the same breath as the earnings beat. A colder fear sits underneath: people who own the AI on one side, people whose work it does on the other, no visible way across. The ground keeps moving. Prompt engineering was a job title. OpenClaw picked up a hundred thousand GitHub stars and caused a Mac Mini shortage in the same week. Ralph was the future for about a weekend. Context engineering, then harness engineering, then graph engineering. Each was the obvious way to work, right up until the next one was. The gaps keep getting shorter. And everywhere, teams ship code faster than they can understand it. Ask whoever’s on call.